The Dialogic Evolution of AI-Based Products: A Polyphonic Analysis of Temporal Transformations
Publication Name
International Journal of Human-Computer Interaction
Abstract
Drawing inspiration from Michail Bakhtin’s philosophy, this study introduces the Polyphonic Model, grounded in the Bakhtinian principles of dialogue and polyphony. These principles emphasise the multiplicity of perspectives and the dynamic, dialogic interplay of various subjective consciousnesses, reflecting the inherent diversity of human thought. This theoretical foundation informs our analytical tool and conceptual framework, enabling an exploration of the recursive and cyclical dialogues between AI and its stakeholders, including AI itself as a non-human participant. Distinctively, the Polyphonic Model’s strength lies not merely in its capacity to highlight the multiplicity of these voices–as many preceding models have–but in its emphasis on the recursive, cyclic nature of the dialogic relationships and a dynamic interchange that continuously reshapes both the AI and its users, especially evident in the temporal transformations. The exploration asserts AI itself as a non-human stakeholder, contributing to and being molded by the very discourses that seek to understand it. While analyzing seventeen diverse stakeholders using ChatGPT as an illustrative context, the exploration zeroes in on two: the non-human stakeholder, exemplified by ChatGPT 4, and its users. Through an innovative methodological approach–an interview with ChatGPT 4 and a sentiment analysis of tweets from its early adopters powered by a lexicon-based approach and a machine learning method, along with Latent Dirichlet Allocation (LDA) for topic modelling–this study examines the dialogic relationships between stakeholders. The findings solidify the Polyphonic Model’s potential not only as an analytical tool but also as a blueprint for the holistic evaluation of AI-powered products. The scope of this study is confined to the development and societal integration of ChatGPT, as a representative example of advanced language processing AI technologies.
Open Access Status
This publication is not available as open access